Research
Publications
-
Does the implementation of external oversight in policing improve public perceptions of police legitimacy? Civilian review boards (CRBs) are frequently promoted as mechanisms to enhance the legitimacy of police agencies by providing independent oversight. Despite public support for CRBs, their adoption and effectiveness remain limited, raising concerns about their actual impact on procedural fairness and police legitimacy. This study assesses the role of CRBs in shaping public perceptions by examining various decision-making scenarios involving police chiefs and CRBs. Using a survey experiment fielded to 2,503 respondents, we investigate whether CRBs enhance legitimacy when they either coincide with or conflict with police chiefs’ determinations in cases of officer misconduct. Our findings suggest that while CRBs may enhance perceptions of procedural fairness for some, particularly those with negative views of police, their involvement does not generally increase legitimacy. In fact, when CRBs conflict with police chiefs, they may diminish public trust in both policing and civilian oversight and further entrench politically polarized attitudes towards policing. These results provide empirical evidence to support concerns that CRBs might not fulfill their intended role in enhancing police legitimacy, especially in cases of institutional disagreement.
-
Every year, Americans elect hundreds of thousands of candidates to local public office, typically in low-attention, nonpartisan races. How do voters evaluate candidates in these sorts of elections? Previous research suggests that, absent party cues, voters rely on a set of heuristic shortcuts—including the candidate’s name, profession, and interest group endorsements—to decide whom to support. In this paper, we suggest that community embeddedness—a candidate’s roots and ties to the community—is particularly salient in these local contests. We present evidence from a conjoint survey experiment on a nationally representative sample of American voters. We estimate the marginal effect on vote share of candidate attributes such as gender, race, age, profession, interest group endorsements, and signals of community embeddedness— specifically homeownership and residency duration. We find that voters, regardless of political party, have strong preferences for community embeddedness. Strikingly, the magnitude of the residency duration effect rivals that of prior political experience.
-
This paper examines the extent to which social pressures can foster greater responsiveness among public officials. I conduct a non-deceptive field experiment on 1400 city executives across all 50 states and measure their level of responsiveness to open records requests. I use two messages to prime social pressure. The first treatment centers on the norm and duty to be responsive to the public’s request for transparency. The second treatment is grounded in the peer effect literature, which suggests that individuals change their behavior in the face of potential social sanctioning and accountability. I find no evidence that mayors are affected by priming the officials’ duty to the public. The mayors who received the peer effects prime were 6–8 percentage points less likely to respond, which suggests a “backfire effect.” This paper contributes to the growing responsiveness literature on the local level and the potential detrimental impact of priming peer effects.
Link to paper here.
NotebookLLM Podcast -
Urban–rural differences in partisan political loyalty are as familiar in the United States as they are in other countries. In this paper, we examine Gallup survey data from the early-2000s through 2018 to understand the urban–rural fissure that has been so noticeable in recent elections. We consider the potential mechanisms of an urban–rural political divide. We suggest that urban and rural dwellers oppose each other because they reside in far apart locations without much interaction and support different political parties because population size structures opinion quite differently in small towns compared with large cities. In particular, we consider the extent to which the compositional characteristics (i.e., race, income, education, etc.) of the individuals living in these locales drives the divide. We find that sizable urban–rural differences persist even after accounting for an array of individual-level characteristics that typically distinguish them.
Link to the paper here.
-
Link to Oxford Bibliographies here.
Other Publications
-
Link here.
Working Papers
-
Given a patchwork system of overlapping local institutions, can residents direct public policy? Current approaches to representation at the local level may present a distorted view of how democracy operates because they fail to account for the overlapping nature of institutions. To address this gap, I first implement a framework that incorporates multiple overlapping governing institutions: cities, counties, school districts, and special districts. Second, I use data from more than 500,000 survey responses to estimate a novel measure of local ideological preferences for cities over time. Finally, to assess the impact of ideology on public policy outcomes, I use a Bayesian within-between random effects model. This methodology yields three major findings. First, I demonstrate that cross-sectional responsiveness exists. Second, I find evidence for dynamic responsiveness in spending but inconclusive evidence for taxation. Third, I provide descriptive evidence that consolidated governance fosters greater responsiveness. I reframe the responsiveness discussion from a single governing unit to a holistic system of overlapping institutions and provide the strongest evidence to date that local governments respond dynamically to the ideology of citizens.
-
Where and why are discriminatory ordinances adopted? Theories of racial threat suppose that members of a racial majority group regard the presence of minorities as a threat to their socio-political status and implement policies to hurt that minority population. I use the racial threat hypothesis to examine the adoption of criminal activity nuisance ordinances (or crime-free housing laws). These ordinances allow officials to designate specific properties and residents as nuisances after repeated police interactions. After that designation, property owners are penalized with fines or the seizure of property if they do not respond by removing the residents. Using data from Ohio municipalities, I find that the racial composition of cities predicts the emergence of criminal activity nuisance ordinances. I attempt to rule out alternative hypotheses surrounding the proportion of renter-occupied housing, crime, and poverty. In further exploring the results, I use a machine learning technique called Random Forests to uncover the discontinuity or “tipping point” where the propensity for adopting such a policy sharply increases or decreases. This research speaks to the generalizability of the racial threat hypothesis, the importance of representation, and the nation's diversification.
-
Regression discontinuity designs are a powerful and increasingly popular tool for causal inference. However, they suffer from a few weaknesses: handling variable data, high sensitivity to noise, and low statistical power. Even when researchers follow best practices, large treatment effect estimates can result from artifacts rather than meaningful discontinuities. These issues undermine the otherwise transparent and assumption-light causal inference that makes RDDs appealing. We introduce Bayesian change point analysis as a principled method to assess discontinuities within the regression discontinuity framework. Change point analysis efficiently identifies and evaluates discontinuities in the raw data without pre-specification. Researchers can use change point analysis to compare detected breakpoints against natural variation in the data and assess whether a discontinuity aligns with theoretical expectations. It strengthens confidence when it detects the expected discontinuity and reveals threats to interpretation, such as noise or anticipation, when it finds others. Through simulations and empirical applications, we demonstrate how change point analysis improves data visualization and validates breakpoints against theory. We provide a sequential workflow for RD analyses and introduce a software package that allows researchers to apply and visualize change point analysis in their own studies.
Slides (Updated July 17, 2024)
Poster: Asian and MENA Polmeth 24
Note: The paper is in development, but if you have comments and suggestions, please send them to bryant.moy@nyu.edu. Thanks!
-
Does ideology operate in similar ways at the national and local levels? The current debate about the existence of responsiveness at the local level rest on the answer to this descriptive question. Previous studies suggest that non-ideological factors are the main drivers of local politics, while more recent research has highlighted the impact of mass/aggregate ideology on local government policies. However, the relationship between ideology and local governance is not well understood, as previous research may have conflated local-level preferences (national views disaggregated to the local level) with local-government preferences (attitudes about cities, counties, and school districts). This short article examines the relevance of self-placement ideology to local politics research. To address this issue, we conducted a survey of Americans to assess how self-placement ideology reflects individual attitudes on local taxation and spending at various levels of government (city, county, school district, and federal). We explicitly target a series of descriptive estimand of the observed ideology of individuals and their support for fiscal policy across levels of government. Our results indicate that self-placement ideology is a measure of residents' general preferences on taxation and spending, applicable across levels of government.
-
Why do some ethnic groups produce local political leaders while others do not? We argue that the spatial distribution of ethnic groups within cities -- particularly their concentration into ethnic enclaves -- shapes political candidate emergence. Ethnic enclaves facilitate leadership by reducing mobilization costs, enabling targeted public goods provision, and fostering dense social and economic networks. Using a novel approach that combines machine learning classification of candidates' ethnic ancestries with spatial measures of ethnic clustering, we analyze data from 638 U.S. cities over five decades. We find that greater geographic clustering significantly increases both the emergence and electoral success of co-ethnic candidates, especially in city council elections. This relationship is nonlinear, intensifying beyond a threshold of spatial concentration. Our findings demonstrate that spatial concentration, beyond simple population share, shapes pathways to local political leadership.
-
Regression discontinuity designs rest on local comparison, yet most studies do not describe the units that determine their estimates. When researchers omit that description, readers cannot judge how far the findings travel. This omission can cause inference to drift beyond the warranted population, stall theory development by obscuring the conditions under which mechanisms operate, and misguide policy advice. We outline a procedure that computes kernel-weighted covariate means for near-cutoff units and compares them with the nominal sample in order to characterize empirical scope conditions. We demonstrate our approach by re-analyzing three published studies using sharp, multi-cutoff, and fuzzy designs. In some cases the analytical sample differs meaningfully from the nominal sample; in others it closely resembles it. Both findings are informative. The procedure makes sample composition explicit, clarifies empirical scope conditions, and strengthens the descriptive foundation on which causal inference rests.
* We are in the process of incorporating comments and feedback. Feel free to email me at Bryant.moy@nyu.edu with additional feedback.
-
Public support for reparations among White Americans remains divided despite evidence that housing discrimination has widened the racial wealth gap. This study examines whether informational and visual interventions that present the structural roots of racial inequality can increase support for reparative policies. We focus on racially restrictive covenants that barred non-White buyers or renters from occupying specific neighborhoods. In a preregistered survey experiment, 4,998 White respondents were randomly assigned to one of five groups: a pure control group, a neutral housing control group, or one of three treatments, including a text only condition, a text condition accompanied by an image of a Black family, and a text condition accompanied by an image of a White family. The text-only treatment increased support for reparations by 7 points on a 0–100 scale, while adding either image resulted in only marginal additional change. Treatment effects were consistent across partisan and ideological groups, including Republicans and Conservatives. These findings have implications for addressing racial injustice even as access to historical accounts of discrimination is being restricted.
-
Black Americans face the highest violent victimization rates and most discriminatory enforcement, yet express greater baseline support for police expansion than White Americans. Increasing police reduces homicides—saving more Black lives in per capita terms—but also increases arrests for minor offenses disparately enforced on Black Americans. Using a pre-registered experiment (N = 3,000 Black and White U.S. adults), we test how citizens weigh policing's double-edged sword. Information about homicide reduction fails to increase support and produces a negative effect among Black respondents. Information about increased arrests for minor offenses reduces support, an effect more than twice as large for Black respondents. When both consequences are presented, the net effect is negative, though protective information partially offsets coercive costs. These results reveal what we call coercive dominance: citizens weight policing's coercive costs far more heavily than its protective benefits, an asymmetry amplified by lived experience with discriminatory enforcement. This poses fundamental challenges for co-producing the public good of safety through police reform in an unequal multiracial democracy.
[PRESENTING AT SPSA: JANUARY 2026]
Research in Progress
-
Citizens resent groups that profit from others' hardship, yet this resentment does not reliably dictate policy preferences. Work on housing attitudes shows that profit-based appeals reduce support for development, but this conflates market actors with opposing interests. Developers profit from building, while landlords profit from scarcity. We argue that profit frames shape attitudes by changing who appears to profit unfairly from policy change versus the status quo. Using a preregistered experiment with 3,000 U.S. adults, we find that developer profit frames reduce support, while frames identifying landlords as beneficiaries of artificial scarcity significantly increase support. These findings demonstrate that disaggregating market actors reveals how economic resentment can mobilize both opposition and support for development, depending on which rent-seeker is targeted.
-
How can scholars obtain efficient difference-in-differences (DiD) estimates when treated cohorts are small, treatment is staggered, and effects are heterogeneous? While recent DiD advances focus on identification, less attention has been paid to estimation—specifically, the bias-variance tradeoff for efficient standard errors. This gap limits scholars working with small treated cohorts. We propose a Bayesian hierarchical model that borrows strength across cohorts in staggered DiD settings, yielding more precise estimates. The model pools more when estimates are noisy and less when they are precise. Simulations show our approach performs favorably relative to both design-based and shrinkage-based DiD estimators, particularly under treatment heterogeneity and dynamic effects. Because of its two-stage construction, our method can be combined with design-based DiD models, allowing researchers to maintain clean identification while gaining estimation efficiency. We demonstrate this combined approach in an empirical application.
-
The mayoral email archive is a novel dataset that includes non-private emails archived from the inboxes and sent boxes of mayors between January 1, 2018, and March 31, 2018. I collected these emails via a series of open records requests sent out in the summer of 2018. The open records requests were initially a part of a project on social pressure and mayoral responsiveness that was published in the Journal of Experimental Political Science in 2021.
I am currently downloading, parsing, and cleaning the email records. After the cleaning process is complete, I will incorporate the emails into my research agenda on local governance.
-
We are currently developing a nonparametric Bayesian approach — Gaussian Process Regressions — to estimate small-area public opinion from large nationally representative surveys.